311: Interdiff includes changes from other changesets
- Fixed
- Review Board
timothys*********@gmai***** (Google Code) (Is this you? Claim this profile.) | |
Oct. 5, 2009 | |
2159, 2954 |
What's the URL of the page containing the problem? http://reviewboard.eng.vmware.com/r/5727/diff/2-4/#1 What steps will reproduce the problem? 1. post-review a changeset 2. submit a separate change that affects a file in the changeset from step 1 3. do another post-review of the changeset from step 1 4. view the interdiff between r1 and r2 What is the expected output? What do you see instead? The interdiff should not include changes that were a part of step 2. What operating system are you using? What browser? Linux. Firefox 2.0 Please provide any additional information below. In the URL at the top, the change to the "Contributors" list was part of another changeset. I realize that it's probably difficult to fix this, but I thought I'd make a note of it.
It took me a bit to understand exactly what was going on here. So for the record, if a diff was put up against a certain file revision in the repository and then later a new diff was put up against a newer revision, the changes between those file revisions will appear in the diff. Yes, this is probably very difficult to track down. I'm unsure of what we can do about it, but there's probably a couple of things.
-
- Priority-Medium + Priority-High + Component-DiffViewer
This is definitely an annoying issue when developers are rebasing their work, but I'm not sure it's going to be possible to fix. We may have to just make sure developers know what will happen when they update their working copy.
-
+ Confirmed
In my case, interdiff is displayed only for the first file, but the rest are diffs against the baseline revision. Unfortunately we cannot give you any real information about the problem since ours is close-sourced. Interdiff is quite a critical feature, at least for us, as it is the key to iterative review process. This seems to be outstanding for a long while, is it so hard to fix?
I believe that another way to reproduce this bug is: 1. Post a diff P1 that includes changes to file A 2. Revert the changes to file A 3. Post an new diff P2 that doesn't contain any changes to A 4. Inter-diff P1 and P2. It will look as if P2 includes the changes from P1 -- that is, P1 and P2 are swapped.
We've thought about this long and hard (since interdiffs were implemented), and there's really no way around it. Alas.
-
- Confirmed + WontFix